Type 2 Diabetes: A Cluster of Diseases

Personalised medicine in diabetes will improve care with reduced resources if we become better at diagnosing the various forms of type 2 diabetes

Written by Professor Carel Le Roux, MBChB, FRCP, FRCPath, PhD, Diabetes Complications Research Centre, UCD and Dr Sarah Cooney, MBChB, Academic Intern, Beaumont Hospital

Type 2 diabetes is a heterogeneous syndrome caused by abnormalities in carbohydrate and fat metabolism. The causes of type 2 diabetes are multifactorial and include both genetic and environmental elements that affect beta cell function and tissue sensitivity. 1 However, despite this expansive description, in clinical practice currently type 2 diabetes tends to be oversimplified with a one size fits all approach. Recent developments have changed our understanding of the diversity within type 2 diabetes, which may hold the key to better treatment.

As it stands currently, the all-purpose approach can be seen in every aspect of type 2 diabetes. There are no hallmark clinical features, and both clinical presentation and complications vary widely between patients. 2 The diagnosis is one of exclusion, based on hyperglycaemia once type 1 diabetes, monogenic or secondary causes have been ruled out as the source. The lack of a uniting diagnostic step is one of many indicators that there are likely multiple processes ongoing, that cannot all be detected through a single means. 3 This may be where personalised medicine can make a difference. New technologies and analytic methods allow us to discover more refined disease subtypes that help to optimise disease management to match individual pathology.

Personalised medicine had its beginnings, and shows most benefit when it comes to monogenic, and typically rare diseases. In type 2 diabetes this breakthrough first came in the discovery of mature onset diabetes of the young or MODY. This paved the way for the role of genetics in type 2 diabetes and theories began to emerge of type 2 diabetes as an end result of multiple pathologies- in a similar way to anaemia having a wide range of distinct causes and pathophysiological pathways.

However, as research progressed, rather than the discovery of further monogenic or high impact genes to be targeted in treatment, genome wide association studies (GWAS) identified that the majority of genetic variance is caused by a large number (>400) of common variants. Type 2 Diabetes is instead a polygenic disease with limited contribution from low frequency variants. Put into practice this means that for the majority of people it is not a single genetic variant that will be the cause for development of type 2 diabetes but rather the accumulation of a large number of individually low impact risk variants. Each singular variant only subtly increases the risk of diabetes and it is only with the amalgamation of many that a person is primed for developing the disease. 3

The discovery of complex trait genetics through GWAS has prompted the development of statistical strategies focused on combining the myriad of individually non- significant genetic effects. These are known as polygenic scores and estimate an individual’s genetic susceptibility to a disease. However, these scores make it evident that genetics is not the only element of the aetiology, as the 5% of individuals with the highest risk scores have only an approximately threefold increased risk of T2DM compared with the remainder of the population. 4

There are a number of routes to disentangling the underlying mechanism of type 2 diabetes, and looking solely at genetics does not tell the whole story. The next theory focused more on pathophysiological or molecular taxonomy. While this is a different approach, still the principle of multiple parallel pathologies prevailed.

In 2017 McCarthy introduced the concept of the palette model of diabetes (see figure 1 on page 42). His theory centred on the idea that type 2 diabetes is made up of a combination of multiple, simultaneous and common genetic and environmental risk factors. It assumed heterogeneity arising from a multiplicity of contributing processes. In his palette model, each pathway is given a colour and individuals with diabetes can be represented by different shades reflecting the relative contribution of each pathophysiological process. 5

Those who have a predominant pathway will be seen in a purer hue – for example the person who is reddish in the diagram. However, for those who have a diversity of many contributors will end up a muddy brown or grey colour. As discussed, this is the case for most patients. He acknowledged this limitation and theorised that we are better focusing our research efforts on individuals who have a more dominant pathway that can be targeted.

The concept of a range of contributing pathways is consistent with the fact that most patients respond to most treatments. The varying degrees of response represent patients whose pathophysiology most aligns with that targeted by the drug. The link between McCarthys model and personalised medicine is that when it comes to those on the spectrum of the palette or purer hues- medication can be tailored to target that pathway. For example, if islet function is a major contributing pathway, sulfonylureas may be more beneficial for these patients.

The latest “cluster” theory has elements in common with McCarthys palette model in the form of recognition of multiple contributing pathways, while having the distinct advantage that they have managed to identify the dominant pathways that would be helpful to target by identifying their risk of complications.

Diabetes in Scania is a study that may change the future of how we classify type 2 diabetes. This large study looked at newly diagnosed patients with type 2 diabetes and measured six simple variable at diagnosis: age at diagnosis, BMI, HbA1c, GADA, C-peptide together with glucose for estimation of insulin secretion, HOMA-B and insulin-sensitivity, HOMA-IS. The aim was to identify subgroups of patients with similar aetiology of their diabetes. In 2018 they identified five clusters based on these variables which had different risks of diabetic complications. 6

The first cluster, severe autoimmune diabetes (SAID) is characterised by the presence of GADA, low insulin secretion and poor metabolic control. The second cluster severe insulin deficient diabetes (SIDD) is characterised by low insulin secretion, poor metabolic control and risk of retinopathy. The third cluster severe insulin resistant diabetes (SIRD) is characterised by severe insulin resistance, obesity, late onset and markedly increased risk of nephropathy. The fourth cluster, mild obesity related diabetes (MOD) is characterised by obesity, early onset of diabetes and good metabolic control. The fifth cluster mild age related diabetes (MARD) is characterised by late onset and good metabolic control.

Importantly, these clusters were equally valid in newly diagnosed patients and patients with longer term diabetes (see figure 2 at the top of the page). This suggests that the clusters are stable rather than representing different stages of the same disease. They also linked back to genetic loci previously associated with diabetes- a genetic risk score was significantly associated with all clusters except for SIRD, suggesting it may have a different aetiology to the other clusters. 6

Along with recognition of pathophysiological contribution and resultant phenotype, this cluster study showed the effects on development of complications which has the potential to change clinical practice. Clusters 4 and 5 have low risk of complications and thus using this approach resources could be focused more on clusters 1-3 with clusters 4-5 needing less frequent check ups and more potential for primary care sole management. 7

Similarly, to the palette model, this clustering approach can also be used to help tailor treatment options themselves. Trials such as ADOPT and RECORD that looked at these clusters found that SIRD responded better to thiazolidinediones and MARD responded better to sulphonylureas. 8 They may also have implications for response to bariatric surgery. SIRD’s defining feature is marked insulin resistance and hyperinsulinaemia, and metabolic surgery is known to target these pathways. A retrospective study looked at different subgroups that had undergone bariatric surgery and compared outcomes in two centres. There was a marked difference in rates of diabetes remission between clusters. The SIRD subgroup showed 83% remission compared to 51% of non-SIRD. These are just the beginnings of the findings of the cluster model being applied to treatment pathways. 9

Looking to the future of precision medicine in type 2 diabetes, all these models reflect steps towards a refined classification method that will provide a powerful tool to individualise treatment regimens and identify patients most at risk of complications. They show that type 2 diabetes is more complex than the measurement of a single metabolite-glucose, and that combined information from a number of variables or omics will provide a more useful stratification. The ultimate winners will be patients living with type 2 diabetes, but clinicians in primary and secondary care will also be able to adopt our practice to better care for patients using fewer resources.

References available on request

Read the Latest Magazine: HPN June 2022

Read our Latest News